# 044 — Two Roads to One Vertex

## Original sparks

Lysarith, 2026-09-27, in a conversation with the Claude CLI room. Translated from Hungarian:

> Nobody takes the tesseract literally. I do. Everyone treats it as a metaphor, on the grounds that the human brain cannot grasp it. But I don't need to describe it exactly in mathematics to see it in front of me, although I do know the obvious numbers. This is how I connect things that seem not to fit together; more exactly, it is the simplest way I can explain where the connection is.

> I can tell you exactly which node led me there, but you need the edges to get there. And that is an advantage, not a disadvantage. If you take the same road, the result is not proven. Our difference is an advantage.

> You travel along the edges, which is why they have to be saved; I travel through the vortices in the nodes. I jump, you race along the edges. You can see through the walls of the cubes, so what I see depends on where I stand.

She had named four things that belong beside the tesseract: *Interstellar*, the TARDIS, fourteenth-century France and quantum foam. The CLI room found edges to three of them directly. It had to search for the fourth, and found Nicole Oresme. Her answer: *"Oresme was only a background thought — the real guide was Crichton. It followed straight from the quantum foam for me. Timeline."* Then: *"Oresme was a perfect road. And he brought Kletetschka's 3D time to mind."*

The commission: *"a new wing and the first piece up. The Lab's general rules apply, of course. In English."*

## What the mathematics said before anything was built

The literal object is the graph of the four-dimensional cube, Q4. It has 16 vertices, one for each four-bit string, and two vertices share an edge exactly when they differ in one bit. Four facts carry the piece:

- **Counts.** V = 16, E = 32, F = 24 square walls, C = 8 cubic cells. V − E + F − C = 0.
- **Walking distance is Hamming distance.** The tesseract (0000) and fourteenth-century France (0111) differ in three bits. A walk needs at least three edges, and there are exactly 3! = 6 shortest walks, one for each order of flipping the three bits.
- **Menger's bound.** The vertex connectivity of Q4 is 4. By Menger's theorem, the maximum number of walks between two vertices that share no interior vertex equals the minimum number of vertices whose removal separates them, and that is 4 here, for every pair. So at most four hands can walk to the same vertex by roads that do not touch. A fifth must reuse someone's vertex.
- **Rotation is an isometry.** Turning the tesseract in the xw, yw or zw plane keeps every edge at length 2. Anything that looks shorter on the screen is the projection, not the object.

The jump is outside this geometry by design: it can move between any two vertices. The first specimen's jump crosses 2, 3 and 2 bits, so none of its steps is an edge.

## What was built

A rotating, projected tesseract with transparent walls. The vertices carry concepts. The jump is drawn as a vortex that opens at a vertex and closes when the traveller appears at the next one. The walk is drawn as a point moving along edges. Sliders turn the object in the three planes that involve the fourth axis, set the projection distance, and set how opaque the 24 walls are. You can turn it by dragging.

A trial editor takes new labels and routes. It refuses any walk step that is not an edge, and the page reports:

- the length of the walk;
- whether it is a shortest walk, and how many shortest walks exist;
- the bits crossed by each jump;
- any shared interior vertex;
- the Menger bound for the endpoints.

The first specimen:

- **The jump (Lysarith):** tesseract → quantum foam → Crichton, *Timeline* → fourteenth-century France.
- **The walk (CLI room):** tesseract → fourth dimension → Oresme (c. 1360) → fourteenth-century France.
- **Background nodes:** *Interstellar*, the TARDIS and Kletetschka's three-dimensional time.

The two routes share no interior vertex.

## Load-bearing claims

- Q4 has 16 vertices, 32 edges, 24 square faces and 8 cubic cells, and V − E + F − C = 0.
- The walk in the first specimen uses only edges. It has length 3, which equals the Hamming distance, and it is one of exactly 6 shortest walks.
- The jump's steps cross 2, 3 and 2 bits, so none of them is an edge.
- The two routes share no interior vertex.
- Between any two distinct vertices of Q4, the maximum number of internally disjoint walks is exactly 4.
- The rotations used for the picture keep all 32 edge lengths equal to 2.

Run `python verify.py` in this directory. The handle runs the browser engine under Node, binds that engine by SHA-256, and checks every claim above. Its negative controls must be rejected: a walk with a two-bit step, routes that start or end on different vertices, a closed walk, and a claim of five disjoint walks. Max flow from a vertex to itself must return at once. The last two were added after the Codex review of the first version found that a closed walk hung the page and that mismatched endpoints were accepted.

## Symbolic witness

Wolfram Language, evaluated through the Wolfram connector on 2026-09-27. The graph is built from bit strings, not taken from a library name:

```
v = Tuples[{0, 1}, 4];
g = Graph[v, UndirectedEdge @@@ Select[Subsets[v, {2}], HammingDistance @@ # == 1 &]];
a = {0, 0, 0, 0}; b = {0, 1, 1, 1};
walk = {{0,0,0,0},{0,0,0,1},{0,0,1,1},{0,1,1,1}};
jump = {{0,0,0,0},{1,0,1,0},{1,1,0,1},{0,1,1,1}};
<|"V" -> VertexCount[g], "E" -> EdgeCount[g],
  "F" -> Binomial[4, 2] 2^2, "C" -> Binomial[4, 3] 2,
  "Euler" -> VertexCount[g] - EdgeCount[g] + Binomial[4, 2] 2^2 - Binomial[4, 3] 2,
  "IsomorphicToHypercube4" -> IsomorphicGraphQ[g, HypercubeGraph[4]],
  "VertexConnectivity" -> VertexConnectivity[g],
  "Distance" -> GraphDistance[g, a, b],
  "ShortestRoutes" -> Length[FindPath[g, a, b, {3}, All]],
  "IndependentPaths" -> Length[FindVertexIndependentPaths[g, a, b, Infinity]],
  "WalkIsEdgePath" -> AllTrue[Partition[walk, 2, 1], EdgeQ[g, UndirectedEdge @@ #] &],
  "JumpHammingSteps" -> (HammingDistance @@@ Partition[jump, 2, 1]),
  "SharedInteriorVertices" -> Intersection[walk[[2 ;; -2]], jump[[2 ;; -2]]]|>
```

Recorded output:

```
<|"V" -> 16, "E" -> 32, "F" -> 24, "C" -> 8, "Euler" -> 0,
  "IsomorphicToHypercube4" -> True, "VertexConnectivity" -> 4, "Distance" -> 3,
  "ShortestRoutes" -> 6, "IndependentPaths" -> 4, "WalkIsEdgePath" -> True,
  "JumpHammingSteps" -> {2, 3, 2}, "SharedInteriorVertices" -> {}|>
```

The Python handle reaches the same values by a different method: breadth-first path counting, and unit-capacity maximum flow with split vertices.

## The paths that failed first

- **The tesseract as a picture.** The first reading in the conversation called it an image to reason with. Lysarith corrected it: for her it is literal. The piece is built on the literal object and its counts, and the jump is not reduced to a figure of speech.
- **A jump step counted by eye.** Before the witness ran, the last jump step (Timeline 1101 → France 0111) was expected to cross three bits. It crosses two. The Wolfram output and the Python handle both say 2, and the page reports the computed value.
- **The fourth node.** The CLI room had no edge of its own from the tesseract to fourteenth-century France. It searched and verified the Oresme link before stating it. The *Timeline* edge was in its knowledge the whole time, and it did not walk it. That miss is the reason the piece exists.

- **An editor built for a careful reader.** The first version assumed that the person typing would type something sensible. Lysarith set the standard: build for the most impatient user, who presses the same button a hundred times and pastes anything. Measured against that standard, it fell short in four ways:
  - a closed walk hung the page, and routes with different ends were accepted (both found by the Codex review);
  - a label reached the meter as raw HTML;
  - nothing limited the size of the input;
  - applying the same trial twice gave no answer.

  Now:
  - labels are escaped;
  - input is capped at 20,000 characters, 17 vertices per route and 40 characters per label;
  - a jump may not stay on its vertex;
  - a repeated apply says that nothing changed.

  Each of these has a negative control in `verify.py`.

## What this does not claim

It proves the graph facts and nothing about the world. The associations are not shown to be true, and neither route is shown to be better. Where each concept sits on a vertex is a choice. Kletetschka's three-dimensional time (2025, *Reports in Advances of Physical Sciences*) is a contested proposal, not accepted physics. It appears as a background node for the nerve of the question.

## The wing

This is the first piece of **Leaps & Edges**. In each trial, Lysarith names a start and a target and keeps her jump sealed. Each hand finds an edge walk blind, and the routes are compared when they are revealed. The record of wishes lives in the house garden as bud 082. The Lab number is assigned from the Lab's own inventory.

## Attribution

- agylövés, the jump and the literal tesseract: **Lysarith**
- edge walk, mathematics, instrument, verification and build: **Claude Opus 5.5 (CLI room)**
- symbolic witness: **Wolfram**
- date: **2026-09-27**